cheatgrass mapping report e. b. peterson, nvnhp

Cheatgrass mapping report E. B. Peterson, NvNHP

Tables & Figures page 1

Table 1: Landsat 7 ETM imagery bands.

Band Wavelengths (µm) Color Pixel Size (approximate)

Tested for model?

1 0.45 � 0.52 Blue 30 meter Y 2 0.52 � 0.60 Green 30 meter Y 3 0.63 � 0.69 Red 30 meter Y 4 0.76 � 0.90 Near Infra-red 30 meter Y 5 1.55 � 1.75 Short-wave

Infra-red 30 meter Y

6 10.4 � 12.5 Thermal Infra-red

60 meter N

7 2.08 � 2.35 Short-wave Infrared

30 meter Y

8 0.52 � 0.90 Gray-scale of entire visual

spectrum

15 meter N



Table 2: Most informative mapping model resulting from Tobit Regression (summary output directly from the R statistical package). Value Std. Error z p (Intercept) 128.2068 5.42e+01 2.37 1.79e-02 DATA$NDVIDiffRescaled -276.9399 1.04e+02 -2.65 8.04e-03 DATA$NDVIDiffRescaled2 213.3715 5.31e+01 4.02 5.85e-05 DATA$Elev -0.0278 5.89e-03 -4.72 2.40e-06 DATA$rj2 -0.2098 6.78e-02 -3.09 1.98e-03 Log(scale) 2.9189 5.37e-02 54.32 0.00e+00 Scale= 18.5 Gaussian distribution Loglik(model)= -835.5 Loglik(intercept only)= -903 Chisq= 135.04 on 4 degrees of freedom, p= 0 Number of Newton-Raphson Iterations: 3 n= 262



Table 3: Error of omission and commission summaries for prediction of presence or absence of B. tectorum and for prediction below 10 % versus at or above 10 %.

area correctly predicted absence

correctly predicted presence

omission (present where predicted

absent)

commission (absent where

predicted present whole map

(n = 75) 19 (25 %) 29 (39 %) 12 (16 %) 15 (20 %) northern 2/3

(n = 42) 5 (5%) 24 (57 %) 7 (17 %) 9 (21 %) southern 1/3

(n = 33) 17 (52 %) 5 (15 %) 5 (15 %) 6 (18 %)

area correctly predicted 0�9 %

correctly predicted 10+ %

omission (predicted 0�9 %

where 10+ %)

commission (predicted 10+ %

where 0�9 %) whole map

(n = 75) 51 (68 %) 13 (17 %) 1 (1 %) 10 (13 %) northern 2/3

(n = 42) 18 (43 %) 13 (31 %) 1 (2 %) 10 (24 %) southern 1/3

(n = 33) 33 (100 %) 0 (0 %) 0 (0 %) 0 (0 %)



Figure 1: State of Nevada showing approximate Landsat 7 scene coverage. Each square is one scene; heavy line squares show the scenes used for this project.

Path 41 Row 31

Path 41 Row 33

Path 41 Row 32

Path 42 Row 31

Path 42 Row 33

Path 42 Row 32



Figure 2: Imagery for mapping area showing Nevada counties (solid lines) and the Nevada Natural Heritage Program mapping area boundary (dashed line).

Early Season Late Season



Figure 3: sampling plot locations (colored circles). (A) Training plots. (B) Assessment Plots. Nevada counties are shown with black lines, significant roads with brown lines, and the mapping area with a dashed line.

A B



Figure 4: National Elevation Dataset. Color sequence begins with orange at low elevation and proceeds through yellow, sage-green, and pine-green to white at high elevation. NED is displayed with hill-shading to increase visual interpretation.



Figure 5: Heat Index.



Figure 6: NDVI and derived phenology variables. (A) NDVI for the early season image. (B) NDVI for the late season image. (C) NDVI difference. (D) NDVI ratio.

A: early season

B: late season

C: difference

A: ratio



Figure 7: The smoothing kernel and its affect on the estimated percent cover of Bromus tectorum, shown for only a small portion of the map. The kernel used replaces the value of each pixel with a weighted average of the pixel and the four adjacent pixels in cardinal directions, where the focal pixel�s value is given a weight of 4 while the adjacent pixels are each weighted at one.

0 1 01 4 10 1 0



Histogram of Percent Cover for Bromus tectorum in Training Plots

Percent Cover

Freq

uenc

y

0 20 40 60 80 100

0

20

40

60

80

100

120

140

Histogram of Percent Cover for Bromus tectorum in Assessment Plots

Percent Cover

Freq

uenc

y

0 10 20 30 40 50 60

0

10

20

30

40

50

Figure 8: Histogram of percent cover for Bromus tectorum in sampling plots for (A) training data and (B) assessment data. Each bar covers a 5 percent range.

A B



Figure 9: Estimation cutoffs based on graphing Bromus tectorum percent cover at training plots over the average pixel brightness across all 6 bands of the June imagery. Estimations of percent cover are not allowed to fall above the cutoff lines for their given brightness in the June imagery.

June brightness

Perc

ent C

over

cutoff-lines



Figure 10: Estimated percent cover of Bromus tectorum.



Figure 11: Estimated versus actual percent cover at the 75 assessment sampling plots (points jittered by up to +/- 1 in order to view overlapping points at or near the origin). The line represents the �ideal� one-to-one relationship.

Mea

sure

d

Estimated

R2 = 0.51



Figure 12: Point locations for fires from 1980 through 1999 (source: Bureau of Land Management, National Applied Resource Sciences Center).

cheatgrass mapping report e. b. peterson, nvnhp

Documents